The earnings tape that crossed at 4:01 PM Eastern on Wednesday was supposed to be the moment Mark Zuckerberg's case for AI capital spending finally cleared. Meta posted revenue of $56.31 billion against a Wall Street consensus of $55.45 billion. Adjusted earnings landed at $7.31 per share, beating the $6.79 estimate. Year-over-year revenue growth came in at 33 percent, the fastest quarter Meta has reported since the post-pandemic boom of 2021.
For about twenty minutes, that was the story.
Then CFO Susan Li opened the prepared remarks with a single revision. Capital expenditures for 2026 will land in the range of $125 billion to $145 billion, up from the $115 to $135 billion range she had given investors three months earlier. The reasons, she said, were higher component pricing plus additional data center costs to support capacity in 2027 and beyond.
Meta shares dropped 7 percent in after-hours trading. By the close on Thursday they were down a further 3 percent, taking the two-session decline to roughly 10 percent and erasing approximately $175 billion in market capitalization. On Thursday morning, JPMorgan analyst Doug Anmuth cut his rating on the stock from Overweight to Neutral and slashed his price target from $825 to $725, a move he tied directly to the new capex profile.
The downgrade tied directly to AI capital intensity at one of the hyperscalers, the kind of analyst action that has been rare in the eighteen months since the 2025 buildout began. And it came on the same night that Alphabet, which had also raised its 2026 capex range, traded up 7 percent.
Wall Street Now Distinguishes Between AI Capex Stories
The juxtaposition was striking. Both Meta and Alphabet reported on Wednesday afternoon. Both beat top and bottom line estimates. Both raised their 2026 AI capex guidance. The market sent one stock up 7 percent and the other down nearly 10.
The difference was not the spending. Alphabet now expects to deploy $180 billion to $190 billion in 2026, up from its prior range of $175 to $185 billion. Google Cloud grew 63 percent year-over-year to $20 billion in revenue, faster than any recent print from Microsoft Azure or Amazon Web Services. Backlog at Google Cloud roughly doubled to $460 billion.
| Company | 2026 Capex Range (New) | Prior Range | Q1 Stock Reaction | Cloud / AI Revenue Signal |
|---|---|---|---|---|
| Alphabet | $180B-$190B | $175B-$185B | +7% | Google Cloud +63% YoY to $20B |
| Meta | $125B-$145B | $115B-$135B | -10% over two days | No direct AI revenue line item |
Meta has no equivalent line item. Its $125 to $145 billion capex commitment funds advertising algorithms, Reels recommendation models, and a pre-training run for whatever comes after the first model out of Meta Superintelligence Labs. The return is supposed to come through better ad targeting and engagement on Facebook, Instagram, and WhatsApp. JPMorgan's downgrade essentially said the math no longer works.
"While we're encouraged by META's +33% Y/Y revenue growth which has been supported by AI-driven ad stack optimizations, accelerating impression growth, and engagement gains, we believe full-stack AI competition is intensifying and Meta has a more challenging path to returns on heavy AI capex beyond advertising." — Doug Anmuth, JPMorgan analyst, in his April 30 downgrade note
Anmuth's model now projects Meta capex jumping another 42 percent to $202 billion by 2027. Under that trajectory, Meta would post negative free cash flow of roughly $4 billion in 2026 and $24 billion in 2027, the first such gap in over a decade.
The Numbers Inside the Number
Meta's actual Q1 spending tells the rest of the story. The company deployed $19.8 billion in capital expenditures during the quarter alone, primarily across servers, data centers, and network infrastructure. Annualized, that pace would already push Meta past the low end of its full-year range.
Zuckerberg told analysts that Meta is now rolling out more than one gigawatt of its own custom silicon designed with Broadcom, alongside a "significant amount" of AMD chips. The intent is to dilute the company's dependence on Nvidia GPUs and bring per-token training costs down before the 2027 capex bill arrives.
He framed the broader thesis in terms that have become familiar from his 2025 letters. "Every sign that we're seeing in our own work and across the industry gives us confidence in this investment," he said on the call, before adding: "We're on track to deliver personal superintelligence to billions of people."
Wall Street has heard the line before. What changed on Wednesday was that Susan Li declined to provide any 2027 capex outlook at all, telling analysts the company is "frankly undergoing a very dynamic planning process" and noting that Meta has "continued to underestimate" its compute needs even as it has ramped capacity. That admission, that the spending bill keeps growing faster than Meta's own forecasts, is what Anmuth seized on.
How Meta's AI Capex Climbed
What This Means for the Rest of Big Tech
Meta is the second hyperscaler in eight days to face a markedly cooler reception on AI spending. On April 24, Microsoft offered voluntary buyouts to 9,000 U.S. employees, the first such program in 51 years, as the company tries to keep operating expense growth contained while infrastructure spending compounds. On April 28, Microsoft formally ended its OpenAI exclusivity, partly because Azure's economics required diversification away from a single model partner.
For ML engineers and infrastructure teams inside Meta, the JPMorgan call carries a specific signal. Meta's planned May 20 layoffs of roughly 8,000 workers were already structured as headcount reduction in service of AI infrastructure. With a $145 billion capex ceiling now formal and a $202 billion 2027 trajectory becoming the new analyst baseline, the financial math will keep pressing on opex. AI Foundations, the Reality Labs cost center, and Llama-era applied research teams are the most exposed.
For practitioners at Anthropic, Google, OpenAI, and the other counterparties Meta buys compute and licenses from, the takeaway is more ambiguous. The capex still gets spent. Meta is not retreating from its 2026 budget. But the next leg of the buildout, the leg that funds 2027 and 2028 compute, is now being underwritten under explicit Wall Street skepticism for the first time. Google's $40 billion deal with Anthropic, announced last week, was structured precisely to shift some of that risk onto a model partner.
The Other Side
Not every Wall Street desk turned bearish on Meta. Several analysts left their Buy ratings unchanged on Thursday and pointed to the same Q1 numbers JPMorgan dismissed. Daily active users on Meta's family of apps grew. Reels monetization continued accelerating. Susan Li reiterated that Meta's 2026 expense growth would land roughly in line with prior expectations.
A Stocktwits roundup on Thursday morning argued that Microsoft, Meta, and Google had collectively committed to $725 billion in 2026 capex across the three companies and that critics of AI spending had been "silenced" by the magnitude of the underlying revenue beats. Under that read, Meta's pullback is a discount, not a verdict.
Zuckerberg's own framing on the call leaned into that defense. He noted that for the first time in Meta's history, the company can build "a first-principles understanding of what you care about and what each piece of content in our system is about." If the bull case is correct and personal-AI ad targeting compounds, the new spending bill arrives looking different in 2027 than it does this week.
The Bottom Line
Meta's Q1 print did everything the company needed it to do. Revenue beat. EPS beat. Growth was the fastest in five years. The first Meta Superintelligence Labs model shipped during the quarter.
None of that mattered for the price. What mattered was a single sentence in a CFO's prepared remarks adding $10 billion to the upper end of an already-historic capex line. Wall Street has spent eighteen months giving hyperscalers a wide pass on AI spending on the assumption that compute would translate into clear revenue. On Wednesday, JPMorgan looked at Meta and decided the assumption no longer held for one of the four companies underwriting the buildout.
The next test arrives in late July, when Meta reports Q2. By then, the company will have executed its May 20 layoffs, shipped its second Superintelligence Labs model, and burned through roughly $40 billion more of the year's capital budget. If revenue growth holds at 33 percent, Anmuth's downgrade becomes the bear-case outlier. If it slows even modestly, $175 billion in lost market value will be the tame estimate.
"Full-stack AI competition is intensifying," Anmuth wrote in the same note where he marked Meta down. The companies that own the full stack — clouds, chips, models, products — get the benefit of the doubt for now. The ones that do not, no longer do.
Sources
- Meta Q1 2026 earnings report (CNBC, April 29, 2026)
- Meta tanks 9%, Alphabet climbs 7% as each company raises capex spend (CNBC, April 30, 2026)
- Meta Platforms gets a downgrade from JPMorgan on massive AI spending forecast (CNBC, April 30, 2026)
- Meta Tumbles 8% on $145 Billion CapEx Bombshell (Yahoo Finance, April 30, 2026)
- Meta just bumped its 2026 capex forecast up to as much as $145 billion (Fortune, April 29, 2026)
- Meta Q1 2026 Earnings Call Transcript (The Motley Fool, April 29, 2026)
- Meta Reports Fourth Quarter and Full Year 2025 Results (Meta Investor Relations, January 28, 2026)
- Alphabet Q1 2026 earnings: Google Cloud revenue up 63% (Yahoo Finance, April 29, 2026)
- Microsoft, Meta And Google Just Silenced AI Spending Critics (Stocktwits, April 30, 2026)
- Meta Stock Loses $175 Billion After AI Expense Estimate Shakes Shareholders (Yahoo Finance, April 30, 2026)